In January 2025, DeepSeek released R1, an open-source reasoning model that has rapidly reshaped the AI landscape. R1 matches GPT-4o-level performance at 90-95% lower compute cost — and it's the first model to successfully integrate web search into a chain-of-thought reasoning framework.
For AEO practitioners, DeepSeek R1 introduces a new AI search surface to track. The model is rapidly being integrated into consumer products across Asia and is already deployed in several enterprise contexts globally.
How R1's search integration differs
Traditional LLM search integration (like ChatGPT with Bing) uses retrieval as a single step: retrieve, then generate. R1's architecture interleaves search with reasoning — it can search, evaluate what it found, decide it needs more information, search again, and continue this loop until it has sufficient context. This means R1's citation behavior is more complex and harder to predict than simpler retrieve-then-generate systems.
DigitAI coverage
We're tracking DeepSeek R1 in DigitAI's Answer Engine Insights. If you're seeing unexpected changes in your AI citation mix, R1's rapid adoption may be a contributing factor. We recommend comparing your brand's visibility across all tracked platforms to identify if any platform-specific gaps have emerged.